A Hybrid of Incremental Technique derived from PLSA
and Standard Language Modelling Techniques for
Auto correction/Word Prediction Applications
This work focuses on the development of an improved
algorithm based on Probabilistic Latent Semantic Analysis
and existing Language Modelling technique(s) that facilitates
language modelling applications such as Auto correction and Word Prediction.
Finding Visual attention of images using hierarchal segmentation approach.
Finding visually attractive region in an image is one of the interesting research
areas due to its real time applications. Generally we take the images to capture
objects (visually attractive region). In the proposed approach regards the important
object detection in an image. The proposed algorithm contains multiple phases:
preparing the super-pixels using ensemble decision trees to estimate the adjacency
probability between the regions, and forming the saliency vector (visually attractive region)
using the learned weights through hierarchal inheritance approach.
Background Subtraction using GPU
Background subtraction (BGS) is a basic task in many computer vision applications,
where we want to segment out the foreground objects from the background of a video.
In the "easiest" case when the camera is static, the background is often defined as
the pixels that stay relatively constant, and the foreground is everything else moving around.
Since BGS is often used as a pre-processing step, it is preferred to be as fast as possible.
Open Domain Question Answering System for Wh-Type Questions
Project is related to devlelopment of system that will answer questions of
Wh-Type ( that are "What","Where","Which",etc) of any domain.
The searching part is done using search engine and after that relevant information
is fetched from top 5 url returned by the search engine